Can customer service AI read the room?
I recently learned about the Korean word 눈치, pronounced nunchi. The closest literal translation is “eye measure,” but in practice it means something in between “emotional intelligence” and “reading the room.” In other words, being really conscious of the context of the conversation you are having, and carefully judging how to act and respond appropriately.
Customer Service Representatives (CSRs), with practice, learn to demonstrate brilliant nunchi. They pick up on a customer’s tone of voice, how they explain what they need, noises in the background (a baby crying, a busy train station etc.) to understand the customer’s situation and use that information to adapt to how they speak to the customer. But it is not always easy: rigid scripts, excessive focus on Average Handling Time, and complex IT systems can get in the way of listening out for these important cues. This can be especially tricky for CSRs working offshore, as some of these non-verbal clues can be culturally specific.
AI as a pair of bionic ears
This is where AI comes into its own. Generative AI is particularly good at listening. This is because it captures everything a customer says, even off-the-cuff remarks, and treats it all with the same level of scrutiny. It can become an extra pair of ears to pick up things that a CSR might have missed.
To illustrate this, I’m going to go back to one of the examples I talked about last week:
Customer has just had a new fibre internet service connected. They have plugged it all in, and nothing seems to be working, even though automated line tests show that the connection should be working (complex)
Customer Service Representative spends 30 minutes running diagnostic tests, getting the customer to reboot the router multiple times, all to no avail. At the end of the call they book an engineer visit to conduct further investigation (complex)
When the engineer visits, she immediately spots that the router had been plugged into the wrong socket (simple)
This is a real situation that happened when I worked for a telecoms company. We were investigating support cases that took longer than expected to resolve, and this was a particularly frustrating example.
One vital piece of information I haven’t told you is what we found out when we listened to a recording of the original call. On that call, the customer repeatedly told the CSR that it was hard to reach their wireless router because it was hidden behind the TV. Meanwhile, written on the CSR’s screen was information that suggested the recently connected internet socket was in the customer’s hallway.
I just wrote this scenario into ChatGPT 4o, and it was able to identify what the CSR had missed:
Generative AI may not be able to demonstrate human levels of nunchi, but it is extremely competent at listening to everything and picking out things that sound unusual. We often think about AI helping CSRs to be more efficient, but the potential for helping them to provide a better quality of service is what gets me more excited.
Recommended news articles
Forbes: Why ‘boring’ AI is the key to better customer service
CX Today: 270,000 Samsung customer support tickets leak onto the internet. Here’s what happened
No Jitter: What agents really think about AI in Customer Service
Japan CX Insider: AI-driven smiles: the answer to a perfect customer experience?
Latest perspectives from BCG
The Future of Field Service with AI
This edition of BCG Executive Perspectives discusses the future of maintenance and field service, and the impact of AI:
What will my team look like? Will I need a different team, or can I upskill?
How will the economics of field services change? What's the ROI on AI tools?
How will the customer experience evolve as a result?
Which tools are best suited, how do I get started…and how do I get this right?